Sound Texture Analysis Based on a Dynamical Systems Model and Empirical Mode Decomposition

This paper describes a system for separating a musical stream into sections having different textural qualities. This system translates several contemporary approaches to video texture analysis, creating a novel approach in the realm of audio and music. We first represent the signal as a set of mode functions by way of the Empirical Mode Decomposition (EMD) technique for time/frequency analysis, before expressing the dynamics of these modes as a linear dynamical system (LDS). We utilize both linear and nonlinear techniques in order to learn the system dynamics, which leads to a successful separation of the audio in time and frequency.